00:32 <whot> Consolatis: static content is unrealistic when you are asked to produce every single page that gitlab might possibly link to, ever

00:32 <whot> Consolatis: you can't disk-space your way out of a git blame request for every git commit in the mesa repository. and you can't cache it either because there's too many and too disparate to rely on caching

00:40 <Consolatis> I agree with git blame and git commits (same for the repo content at each individual commit). Those are likely better served with requiring a user account. git blame I would remove completely. I was mostly talking about Issues and MRs and their index lists

05:33 <daniels> Consolatis: when you say ‘instead of one giant query, I would aggressively cache the content in Redis through many smaller queries so clients are effectively served static content’, you’re describing exactly how GitLab actually works

05:36 <daniels> but yeah, you might want to go read up on LWN’s experience of the crawlers, as a largely-static website with much much much less content to attempt to cache

05:37 <daniels> and then think about the number of URLs on the internet, and whether if you were trying to make an LLM crawler and didn’t care about externalities, whether you’d bother indexing every URL you’d ever seen and querying whether or not you needed to do it again, or just suck up the content en masse and not bother with the URL tracking

08:55 <karolherbst> can somebody point me out where the spam bot fetches emojis? I'd be interested in maybe making the bot a bit more competent in that regard. We've seen a few insulting emoji reactions being used and were wondering if the bot can help out point out if individuals get harassed or targeted by rando accounts or something

08:56 <karolherbst> and insulting/racist emoji reactions is something we (and also other communities) have seen happening in those cases

08:57 <karolherbst> no idea what to do about it yet, but I was wondering digging through the API and see what could be done there

08:58 <karolherbst> those reactions can be removed through the API, but it's a real pain and takes a lot of time doing it all manually